Scribd
Upload
33 views10 pages

Dense Extreme Inception Network

The paper proposes a robust CNN model called DexiNed for edge detection. It was trained from scratch on a new dataset called BIPED containing carefully annotated edges. DexiNed achieves results comparable to other deep learning approaches for edge detection but the evaluation metrics penalize some of the edges it detects.

Uploaded by

Mursal Rehman
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
33 views10 pages

Dense Extreme Inception Network

The paper proposes a robust CNN model called DexiNed for edge detection. It was trained from scratch on a new dataset called BIPED containing carefully annotated edges. DexiNed achieves results comparable to other deep learning approaches for edge detection but the evaluation metrics penalize some of the edges it detects.

Uploaded by

Mursal Rehman
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

DENSE EXTREME

GROUP MEMBERS: INCEPTION


MURSAL JUNAID
REHMAN 22100158 NETWORK:
SARVECH QADIR TOWARDS A ROBUST
22100110
CNN MODEL FOR
EDGE DETECTION
WHAT IS THE PROBLEM?

• Previous deep learning-based


approaches to detect edges trained on
already existing boundary detection or
object segmentation datasets.
• Datasets contain missing edges so after
training the predicted edge-maps lost
some edges in the images.
• Currently, these datasets are used just
for qualitative comparisons.
WHY IS IT IMPORTANT?

• Edge detection is a recurrent task required for several


classical computer vision.
• Used in modern tasks such as image-to-image
translation and photo sketching.
• Also used in fields such as medical image analysis or
remote sensing.
AIM OF THIS PAPER

• A dataset with carefully annotated edges has been generated and released to the
community—BIPED: Barcelona Images for Perceptual Edge Detection.
• A robust CNN architecture for edge detection is proposed, referred to as DexiNed: Dense
Extreme Inception Network for Edge Detection. The model has been trained from the
scratch, without pretrained weights.
OVERVIEW OF BACKGROUND LITERATURE

Brain-biologically
Low level feature
inspiration

Classical learning Deep learning


algorithms algorithms
METHOD USED

Architecture of a Dense Extreme Inception Network

Upsampling block (UB)

Regression loss function

Dataset and annotations from BIPED images dataset is used


RESULTS

• Results are based on F-measure.


• The three versions of DexiNed get the
similar results
• DexiNed-dc upsampling results evaluated.
• The three diagrams are:
(a) DexiNed upsampling versions.
(b)Outputs of DexiNed in testing stage.
© Comparison of DexiNed with other
DL based edge detectors.
SHORTCOMINGS

• Not enough training dataset.


• Evaluation metrics penalize edges detected by DexiNed.
• The results are penalized as datasets are intended for objects’ boundary detection or
objects segmentation.
SUGGESTED IMPROVEMENTS

• Dataset for edge detection is shared.


• Contour and boundary detection problems
• Edge detection and contour/boundary detection are different problems.
Testing
Results from the proposed approach using different
datasets (note that DexiNed has been trained just with
BIPED.

You might also like